JPWO2018051568A1 - Plant abnormality diagnosis device and plant abnormality diagnosis system - Google Patents

Abstract

A plant abnormality diagnosis apparatus and a plant abnormality diagnosis system capable of early judging an abnormality in a whole plant in plant abnormality diagnosis using ART. The plant abnormality diagnosis device 2 applies a plurality of measurement data from various sensors installed in the plant 5 to be diagnosed to data belonging to a category determined by data at normal time according to Adaptive Resonance Theory (ART) and It comprises an abnormality degree calculation unit 103 that calculates the abnormality degree A of the entire plant 5 to be diagnosed based on the difference in the spatial distance from the plurality of measurement data.

Description

  The present invention relates to an abnormality diagnosis apparatus and plant abnormality diagnosis system for a plant using adaptive resonance theory (ART) which is one of clustering techniques.

  Early detection of an abnormal state of the plant before an alarm is output is effective in improving the operation reliability. As a method of detecting an abnormality of a plant early, it is effective to learn in advance a data pattern at the time of normality of the plant and detect a state change by comparing this with a pattern of measurement data. One of such data analysis methods is Adaptive Resonance Theory (ART, hereinafter referred to as ART). ART is one of clustering analysis methods, and is a method of classifying input data into a plurality of categories (clusters). A category represents a group of data having similarity. In a general clustering analysis method, a plurality of input data are mapped on a multidimensional space, and data grouping is defined from the spatial distance. In ART, a category number is assigned to input data. Input data with the same category number indicates that the similarity is high.

When performing an abnormality diagnosis of a plant using ART, first, plant data at the normal time is set as an input of ART, and a category number is output. The category number output at this time is defined as a normal state. The above process is called "learning" in a normal state. If the operating conditions of the plant are different even during normal times, category numbers are defined according to each. Therefore, there are a plurality of category numbers indicating normality.
Next, measurement data of the plant is set as an input of ART, and a category number is output. This process is called "diagnosis". If the category number output in the diagnostic process is included in the category number output in the learning process, it is determined that the condition is normal. On the other hand, when the category number not output in the learning process is newly output, the input data is a data pattern not included in the normal data, and it is determined that the input data is abnormal.

  For example, a technique described in Patent Document 1 is known as one that performs abnormality diagnosis of a plant using such an ART. The plant monitoring support system disclosed in Patent Document 1 detects an abnormality based on information of a category determined by ART using measurement data of a desalination plant, and further evaluates the degree of abnormality and changes the state to an abnormal state. We also select signals with high contribution rates. In order to evaluate the degree of abnormality, an adjustment parameter 利用 for adjusting the size of the category is used. The larger the adjustment parameter ρ, the smaller the category, and the input data can be classified finely. Therefore, even if the data pattern is slightly different from the normal data pattern, it is determined as a new category, so that sensitivity to abnormality can be increased. However, since it becomes easy to determine that an ordinary fluctuation included in input data is abnormal, erroneous detection is increased. On the other hand, the smaller the adjustment parameter ρ, the larger the category and the coarser the classification of the data. Therefore, although the sensitivity to abnormality is lowered, false detection can be reduced.

Under a plurality of conditions in which the value of the adjustment parameter ρ is changed, normality / abnormality is determined by ART, and the degree of abnormality is determined according to the value of ρ determined to be abnormal. For example, when ρ is small, that is, even when the sensitivity to abnormality is low, if it is determined that ART is abnormal, it can be determined that the difference from normal is very large. Using this, the alarm levels such as "large", "medium" and "small" are evaluated.
Moreover, in patent document 1, in order to select the signal with a high contribution rate to an abnormal state, the positional relationship when mapping input data to multidimensional space is used. In the diagnostic processing, if the input data is not included in the category indicating normal, the positional relationship between the input data in space and the center of gravity of the category closest to it is in the direction of each axis (each signal). The contribution rate of each signal is determined from the distance difference.

Unexamined-Japanese-Patent No. 2016-81482

  However, in the configuration described in Patent Document 1, category determination processing is performed each time measurement data of a desalination plant, that is, measurement data by each sensor installed in the desalination plant is acquired, and measurement data from each sensor The contribution rate is calculated for each (every input signal), and no consideration is given to the point at which an abnormality in the entire plant is determined early. In other words, in the configuration described in Patent Document 1, category determination processing by ART is performed each time measurement data obtained by each sensor is acquired. Since the category determination process involves convergence calculation, it takes time for the process, and a time delay may occur before detecting an abnormality, and it is feared that a sudden abnormality in the whole plant is overlooked.

  Therefore, the present invention provides a plant abnormality diagnosis device and a plant abnormality diagnosis system capable of early judging an abnormality in the whole plant in the abnormality diagnosis of a plant using ART.

In order to solve the above problems, the plant abnormality diagnosis device of the present invention is data belonging to a category determined by data at normal times by adaptive resonance theory with respect to a plurality of measurement data from various sensors installed in a plant to be diagnosed. And a plurality of measurement data, and an abnormality degree calculator configured to calculate an abnormality degree of the entire plant to be diagnosed.
In the plant abnormality diagnosis system according to the present invention, a plant abnormality diagnosis apparatus that diagnoses an abnormality of a plant to be diagnosed, a plurality of process control of the plant to be diagnosed and a plurality of sensors installed in the plant to be diagnosed. A plant control apparatus for transmitting measurement data of the plant abnormality diagnosis apparatus to the plant abnormality diagnosis apparatus, and an input / output apparatus having an input unit and a display unit, and the plant abnormality diagnosis apparatus is configured from various sensors transmitted from the plant control apparatus. Of the plurality of measurement data, the abnormality degree of the whole plant to be diagnosed on the basis of the difference of the spatial distance between the data belonging to the category determined by the data at normal time by the adaptive resonance theory and the plurality of measurement data. It is characterized in that it has an abnormality degree calculation unit to be calculated.

According to the present invention, it is possible to provide a plant abnormality diagnosis apparatus and a plant abnormality diagnosis system capable of early judging an abnormality in the entire plant in the abnormality diagnosis of a plant using ART.
Problems, configurations, and effects other than those described above will be apparent from the description of the embodiments below.

BRIEF DESCRIPTION OF THE DRAWINGS It is a functional block diagram which shows the whole schematic structure of the plant abnormality diagnosis system of Example 1 which concerns on one Example of this invention. It is explanatory drawing of the concept of clustering. It is a block diagram which shows the principal part of the ART process part shown in FIG. It is a figure which shows the data structure of the category database shown in FIG. It is a figure which shows the data structure of the plant database shown in FIG. It is a figure which shows the data structure of the diagnostic result database shown in FIG. It is a figure which shows the data structure of the abnormal event database shown in FIG. It is a figure which shows the calculation method of abnormality degree. It is a figure which shows the time change of an abnormality degree. It is a figure which shows the data structure of a diagnostic result database. It is explanatory drawing of the concept of clustering by ART. It is an example of a display screen which displays the time change of an abnormal degree, and the time change of a category number. It is an example of a display screen which displays the time change of an abnormal degree, and the time change of a category number. It is an example of a display screen at the time of registration of an abnormal event. It is an example of a display screen which displays the time change of a category number, and the abnormal event as a diagnostic result.

The plant abnormality diagnostic apparatus and plant abnormality diagnosis system according to the embodiment of the present invention are applied to a seawater desalination plant, a sewage treatment plant, a water treatment plant such as a water treatment plant, a power plant, a chemical plant, etc. Including various industrial plants.
Hereinafter, embodiments of the present invention will be described with reference to the drawings.

  FIG. 1 is a functional block diagram showing an overall schematic configuration of a plant abnormality diagnosis system according to a first embodiment of the present invention. As shown in FIG. 1, the plant abnormality diagnosis system 1 includes a plant abnormality diagnosis device 2 that diagnoses an abnormality of the plant 5, a plant control device 3 that monitors the plant 5 and performs process control, and an input unit and an output unit. It comprises an input / output device 4.

  The input / output device 4 has an input device such as a keyboard or a mouse, for example, and is used when the user inputs and / or registers data to the plant abnormality diagnosis device 2. Also, the input / output device 4 is an output unit having a display device such as a liquid crystal display (LCD) or an organic EL display, and a print output device such as a printer. The diagnosis result of the abnormality or a screen for interactive processing for the user of the plant abnormality diagnosis apparatus 2 is displayed.

  The plant control device 3 is realized by, for example, a control panel or SCADA (Supervisory Control And Data Acquisition). When SCADA is used as the plant control device 3, the SCADA is installed at least in a programmable logic controller (PLC) that performs process control of various devices constituting the plant 5, various devices and piping constituting the plant 5, etc. It has a communication function of collecting measurement data from various sensors such as flow meters, pressure gauges, thermometers, etc., and transmitting measurement data from the collected various sensors to the plant abnormality diagnosis device 2.

  The plant abnormality diagnosis apparatus 2 includes a data acquisition unit 101, a data preprocessing unit 102, an abnormality degree calculation unit 103, an ART control unit 104, an ART processing unit 105, an abnormality determination unit 106, an input / output control unit 107, a plant database 108, and categories. A database 109, a diagnosis result database 110, an abnormal event database 111, and a communication I / F 112 for receiving measurement data from various sensors transmitted from the plant control device 3 are mutually accessible by the internal bus 113 It is connected. The data acquisition unit 101, the data preprocessing unit 102, the abnormality degree calculation unit 103, the ART control unit 104, the ART processing unit 105, the abnormality determination unit 106, and the input / output control unit 107, which constitute the plant abnormality diagnosis device 2, A processor such as a CPU (Central Processing Unit) (not shown), a ROM storing various programs, a RAM temporarily storing data of calculation processes, and a storage device such as an external storage device Reads and executes various programs stored in the ROM, and stores the calculation result as the execution result in the RAM or an external storage device.

The data acquisition unit 101 acquires, via the internal bus 113, measurement data from various sensors received from the plant control device 3 via the communication I / F 112. For example, A / D conversion processing, smoothing processing (Noise removal) or normalization processing is performed. Further, the data acquisition unit 101 transfers measurement data obtained by the various sensors after the above-described processing to the plant database 108 via the internal bus 113. Thereby, the plant database 108 stores data of the plant 5 which is measurement data by various sensors in time series.
The data pre-processing unit 102 accesses the plant database 108 via the internal bus 113, and from the data stored in the plant database 108 takes in data serving as an input of a diagnostic process by ART described later, and further, inputs an ART process Normalize the data between 0 and 1 so that The data preprocessing unit 102 also transfers the normalized data to the abnormality degree calculating unit 103 via the internal bus 113.

The abnormality degree calculation unit 103 takes in the normalized data transferred from the data preprocessing unit 102 via the internal bus 113, and calculates the degree of abnormality in the entire plant 5 to be diagnosed and the degree of abnormality contribution of each signal. calculate. In addition, each signal corresponds to the measurement data by the above-mentioned various sensors contained in the data after normalization here.
The ART control unit 104 takes in the abnormality degree calculated by the abnormality degree calculation unit 103 via the internal bus 113, and controls the execution timing of the process by the ART according to the output tendency of the abnormality degree.

  The ART processing unit 105 executes processing by the ART. The abnormality determination unit 106 takes in the information of the category which is the output result of the ART processing unit 105 via the internal bus 113, and determines the normality or the abnormality based on the information of the category. Further, the abnormality determination unit 106 transfers the determination result of normality or abnormality to the diagnosis result database 110 via the internal bus 113. Thereby, the diagnosis result database 110 stores the determination result of normality or abnormality transferred from the abnormality determination unit 106 via the internal bus 113 as a diagnosis result of the plant 5. The category database 109 stores characteristic data for each category used in the processing of the abnormality degree calculation unit 103 and the abnormality judgment unit 106.

  In the plant abnormality diagnosis device 2, as described above, the ART is used for the diagnosis process of the plant 5 to be diagnosed. Therefore, first, processing by ART executed by the ART processing unit 105, determination of ART processing result executed by the abnormality determination unit 106, and output processing of the diagnosis result database 110 of diagnosis result will be described.

ART is one of the clustering methods. FIG. 2 illustrates the concept of clustering. The upper diagram of FIG. 2 shows the state in which the values of two types of signals (signal X1 and signal X2) are plotted in a two-dimensional space, that is, the correlation between these two signals (signal X1 and signal X2). As shown in the upper part of FIG. 2, the signal X1 on the horizontal axis and the signal X2 on the vertical axis are standardized by the data preprocessing unit 102 between each signal (data) from 0 to 1.
In clustering, as shown in the upper diagram of FIG. 2, data close to each other are put together to define groups (clusters) with respect to data plotted on space. In the example shown in the upper part of FIG. 2, circles 210 a and 210 b each represent one group. In ART, each group is called a category, and different category numbers are assigned to the category 210a and the category 210b, respectively. Therefore, the output of ART is the category number assigned to the input data. Data assigned the same category number indicates that they tend to be similar. In the example shown in the upper drawing of FIG. 2, the two-dimensional case, that is, the case where two signals (signal X1 and signal X2) are input is shown, but this is for the purpose of making the description easy to understand. . In the abnormality diagnosis of the actual plant 5, a large number of signals are set as inputs, and clustering processing is performed in a multi-dimensional space.

  When performing abnormality diagnosis of the plant 5 by ART, first, before the diagnosis processing, category processing is performed by the ART processing unit 105 using plant data defined as normal (measurement data by the various sensors described above) as an input. This process is called learning. The category number output in the learning process represents normal. Next, the plant data for diagnosis is set to the input of the ART processing unit 105 to perform category determination processing, and the abnormality determination unit 106 performs normal / abnormality determination processing based on the category output from the ART processing unit 105. Do. As shown in the lower part of FIG. 2, when the input data 211 deviates from the category 210a and the category 210b defined as normal, the plant 5 is in a state different from the characteristics at the time of normal, and it can be judged as abnormal. In the processing by ART, when data not corresponding to the existing category is input, a category is newly created and a number is assigned. That is, when the category number determined by the ART processing unit 105 is different from the category number at the time of learning, the abnormality determination unit 106 determines that the plant 5 is in an abnormal state. In addition, when the plant 5 is a plant of a new installation, the above-mentioned learning process is performed at the time of test operation of the plant 5. Further, when the plant abnormality diagnosis device 2 is newly incorporated into the existing plant 5, learning may be performed using past operation result data (plant data).

Next, details of clustering processing by the ART processing unit 105 will be described. FIG. 3 is a block diagram showing the main part of the ART processing unit 105 shown in FIG.
As shown in FIG. 3, the ART processing unit 105 includes at least a normalization / noise removal unit (F0 layer) 310, an input pattern holding unit (F1 layer) 311, a category output unit (F2 layer) 312, and a category validity determination. A component (Orienting Subsystem) 313 and a memory 314 are included. The algorithm (processing by ART) executed by the ART processing unit 105 is the following steps 1 to 6.

  Step 1: In the normalization / noise removal unit (F0 layer) 310, the magnitude of the input vector is normalized to 1 and noise is removed.

  Step 2: The input pattern holding unit (F1 layer) 311 performs short-term storage of data using a recurrence equation (also referred to as a difference equation).

  Step 3: In the category output unit (F2 layer) 312, category candidates are selected by comparing the input data that has been input with the weighting factors stored in the memory 314.

  Step 4: In the category validity judgment unit (Orienting Subsystem) 313, the validity of the category selected in the category output unit (F2 layer) 312 is evaluated. If it is determined that the input data is classified, the input data is classified into the category, and the process proceeds to step 6. If it is not determined that the category is determined to be valid, the category is reset (excluded from category candidates), and the process of selecting candidate category numbers from step 2 is performed again.

  Step 5: In step 4, if it is determined that all existing categories are reset (excluded from category candidates), that is, no valid categories exist, create a new category. At this time, weighting factors corresponding to the new category are also newly defined. The newly defined weighting factors are stored in the memory 314.

Step 6: When the input data is classified into categories, update weighting factors corresponding to the corresponding categories. Updating of the weighting factor is performed by the following equation (1).
WJ (new) = Kw x p + (1-Kw) x WJ (old) ... (1)
Here, WJ (new) is a weighting factor corresponding to category J, WJ (old) is a past weighting factor, p is input data (or data derived from input data), Kw is a learning rate parameter, and input is Determine how much data will be reflected in the new weighting factors.

The above is the flow of the clustering process by the ART processing unit 105. The ART processing unit 105 outputs a category number corresponding to the input data by this processing.
The feature of the clustering process by the ART processing unit 105 is the process of step 5. By the process of step 5, when the input data has a tendency different from existing categories (clusters), new categories can be newly created without changing the existing categories. For this reason, it is possible to create a category having new data trends while preserving categories learned in the past.

  A category number corresponding to input data is output by clustering processing by the above-described ART processing unit 105, and the abnormality determination unit 106 determines whether the category number output by the ART processing unit 105 is normal or abnormal.

  The plant abnormality diagnosis device 2 executes the learning process using normal data in advance before performing the diagnosis process. Information on normal categories defined in the learning process is stored in the category database 109. FIG. 4 shows the data structure of the category database 109 shown in FIG. As shown in FIG. 4, the category database 109 includes a "category number" column, a "normal / abnormal" column indicating normal or abnormal, a "number of data" column indicating the total number of data for each category number, and a plant 5 An “input data average value” column indicating an average value for each signal which is measurement data from various sensors installed in various devices and pipes is stored in a table format. That is, for each category number, a flag representing normality / abnormality as 0/1, the corresponding number of input data, and an average value of the input data are stored. For example, if "category number" is "0", "0" indicates the normal category defined in the learning process in the "normal / abnormal" column, "40" in the "number of data" column, The “input data average value” column stores “0.6” which is an average value of “signal A” and “0.3” which is an average value of “signal B”. In addition, if the "category number" is "5", the "normal / abnormal" column does not correspond to the normal category defined in the learning process and "1" indicates that it is abnormal, and the "number of data" column In the “4” and “input data average value” fields, “0.4” which is an average value of “signal A” and “0.8” which is an average value of “signal B” are stored.

  Here, the plant database 108 will be described. FIG. 5 is a diagram showing the data structure of the plant database 108 shown in FIG. In FIG. 5, the case of a power plant is shown as an example of the plant 5. As shown in FIG. 5, the plant database 108 has a table format of “date and time” column, “generator output (MW) (signal A)” column, and “fuel flow rate (kg / s) (signal B)” column. Is stored. In addition to this, although not shown, for example, a plurality of data item fields measured by various sensors such as a "gas temperature" field and a "gas flow rate" field are provided. As shown in FIG. 5, for example, when the "date and time" is "June 1, 2016 00:00", "1000" in the "generator output (MW) (signal A)" column, and "fuel flow rate ( "100" is stored in the "kg / s) (signal B)" column. In addition, when “Date and time” is “June 1, 2016 01:30,” “Generator output (MW) (signal A)” column is “950”, and “fuel flow rate (kg / s) (signal B) “95” is stored in the) field.

  On the other hand, at the time of diagnosis processing, the above-mentioned abnormality determination unit 106 determines normality / abnormality based on the data stored in the category database 109 for the category number output by the ART processing unit 105. That is, the abnormality determination unit 106 searches the category database 109 using the category number acquired from the ART processing unit 105 via the internal bus 113 as a search key, and the normal / abnormal state is determined depending on whether the category is registered as a normal category. I can make a decision. The abnormality determination unit 106 determines that a category number not registered as normal in the category database 109 is abnormal, and stores this information in the category database 109. In the example shown in FIG. 4 described above, the “category number” is determined to be abnormal as “5”. At this time, “1” is stored in the “normal / abnormal” column. In addition, abnormality determination unit 106 calculates an average value of input data for each category for each signal, and outputs the result to category database 109 via internal bus 113. When the CAT processor 105 determines that the "category number" is "5" during diagnosis processing, the abnormality determination unit 106 determines that the "category number" stored in the category database 109 is "5". For "data", while increasing the value in the "number of data" column, "average value of input data" is updated.

  Further, the abnormality determination unit 106 stores the determination results of the category in the diagnosis result database 110 in time series via the internal bus 113. FIG. 6 is a view showing a data structure of the diagnosis result database 110 shown in FIG. As shown in FIG. 6, the diagnosis result database 110 stores "date and time" column, "normal / abnormal" column, and "category number" column in a table format. For example, if the date / time is "June 1, 2016 00:00", the "normal / abnormal" column shows "0" as a result of being judged as normal by the abnormality judgment unit 106, and the "category number" column. The category number “0” output by the ART processing unit 105 is stored. In addition, when the “date and time” is “June 1, 2016 01:30”, the “normal / abnormal” column is “1” in the “category number” column as a result determined as abnormal by the abnormality determination unit 106. Stores “10” which is a category number output by the ART processing unit 105.

  The input / output control unit 107 (FIG. 1) constituting the plant abnormality diagnosis device 2 outputs the data stored in the above-described diagnosis result database 110 and the category database 109 to the input / output device 4 as a diagnosis result. The display form (display screen example) of the display device constituting the input / output device 4 will be described later. Further, the input / output control unit 107 takes in data relating to an abnormal event input by the user via the input / output device 4. When the plant abnormality diagnosis system 1 outputs that an abnormality has occurred in the plant 5 and the cause of the abnormality is determined later, this can be registered in the plant abnormality diagnosis system 1. The data of the cause of abnormality input by the user via the input / output device 4 is stored in the abnormality event database 111 via the input / output control unit 107 and the internal bus 113.

  FIG. 7 is a view showing a data structure of the abnormal event database 111 shown in FIG. As shown in FIG. 7, the abnormal event database 111 stores a "category number" column and an "abnormal event" column in a table format. That is, in the abnormal event database 111, data in text format related to the abnormal event registered by the user via the input / output device 4 is stored in association with the category number output by the ART processing unit 105. For example, when "category number" is "5", "pump failure" is stored in the "abnormal event" column, and when "category number" is "10", "heat exchanger heat transfer efficiency" in the "abnormal event" column. "Degraded" is stored. The input / output control unit 107 accesses the abnormal event database 111 via the internal bus 113 when the above-described abnormality determination unit 106 determines that it is abnormal, and the abnormal event generated in the past stored in the abnormal event database 111 Search for the same thing. The category number is used as a search key for the search. If the abnormal event database 111 contains the same category number as that output from the ART processing unit 105, the data (text format data) of the corresponding abnormal event is read and output to the input / output device 4. By this processing, the plant abnormality diagnosis system 1 not only determines the normality / abnormality of the plant 5, but also provides the user with the information (content of the abnormal event) if it is the same as the abnormal event that occurred in the past. can do.

  The above is the flow of diagnosis processing by ART, but in the plant abnormality diagnosis system 1 of the present embodiment, the above-described processing by ART executed by the ART processing unit 105 and plant diagnosis processing by the abnormality determination unit 106 are fixed. The processing is performed only when the abnormality degree calculated by the abnormality degree calculating unit 103 described later satisfies the execution condition, instead of periodically executing at time intervals of. The category determination processing by the ART processing unit 105 requires a calculation load in order to perform convergence calculation. Therefore, the processing by the ART processing unit 105 is not executed every step, and the ART control unit 104 activates the ART processing unit 105 when it is determined that the state suitable for the category determination processing by the ART processing unit 105 is reached. Do.

  First, a method of calculating the degree of abnormality by the abnormality degree calculator 103 will be described. FIG. 8 is a diagram showing a method of calculating the degree of abnormality. In the upper diagram of FIG. 8, in order to facilitate the description, a case is shown in which categories are created in a two-dimensional space with two signals (signals X1 and X2) as inputs. Here, it is assumed that normal category 1 and normal category 2 representing normal are created by learning processing. Next, when the input data 212 is received via the communication I / F 112 during diagnosis processing, the abnormality degree is defined as the distance from the center of gravity of the closest category. In the example shown in the upper part of FIG. 8, the normal category 1 is closer to the input data 212 mapped on the two-dimensional space than the normal category 2. For this reason, the abnormality degree calculation unit 103 calculates the abnormality degree of the entire plant 5 to be diagnosed, based on the distance to the input data 212 based on the center of gravity of the normal category 1. Further, the degree of abnormality of each signal (signal X1, signal X2) can also be calculated from the positional relationship between the input data 212 and the normal category in the two-dimensional space.

  The lower part of FIG. 8 shows an enlarged view of the centroid X of the normal category 1 which is the closest category in the upper part of FIG. 8 and the plot points of the input data 212. At this time, the distance difference between each axial direction, that is, the axial direction of the signal X1 and the axial direction of the signal X2 represents the difference from the normal state of each signal (signal X1, signal X2). Therefore, in the lower part of FIG. 8, the distance difference ΔX1 in the axial direction of the center of gravity X of the normal category 1 which is the closest category and the signal X1 of the input data 212, and the center of gravity X of the normal category 1 which is the closest category and the input Based on the distance difference ΔX2 in the axial direction of the signal X2 of the data 212, the abnormality contribution Rn of each signal (signal X1, signal X2) is determined using the following equation (2).

よって、信号Ｘ１の異常寄与度Ｒ_Ｘ１は、ΔＸ１／（ΔＸ１＋ΔＸ２）として求められ、信号Ｘ２の異常寄与度Ｒ_Ｘ２は、ΔＸ２／（ΔＸ１＋ΔＸ２）として求められる。

Therefore, abnormal contribution _{R X1} of the signal X1 is determined as ΔX1 / (ΔX1 + ΔX2), abnormal contribution _{R X2} of the signal X2 is determined as ΔX2 / (ΔX1 + ΔX2).

Moreover, when the abnormality degree of the whole plant 5 which is the above-mentioned diagnostic object is A, the abnormality degree Sn of the signal n is calculated using the following equation (3) using the abnormality contribution degree Rn.
Sn = A × Rn (3)
Therefore, the degree of abnormality _{S X1} of the signal X1 is (A × .DELTA.X1) obtained as / (ΔX1 + ΔX2), abnormal degree _{S X2} of the signal X2 is determined as (A × ΔX2) / (ΔX1 + ΔX2). The value obtained by totaling the degree of abnormality Sn of each signal has a relationship equal to the degree of abnormality A of the entire plant 5 to be diagnosed. Therefore, the abnormality degree calculated by the abnormality degree calculator 103 can be displayed in a display form as shown in FIG. FIG. 9 is a diagram showing temporal changes in the degree of abnormality, with the degree of abnormality taken on the vertical axis, the time taken on the horizontal axis, and the degree of abnormality Sn of each signal (signal X1, signal X2) colored There is. As described above, the sum of the abnormality degree (S _X1 ) of the signal _X1 and the abnormality degree (S _X2 ) of the signal X2 becomes the abnormality degree A of the entire plant 5 to be diagnosed. That is, the abnormality degree Sn of each signal (signal X1 and signal X2) can be evaluated as a breakdown of each signal (signal X1 and signal X2) with respect to the abnormality degree A of the whole plant 5 to be diagnosed. With this display mode, when the value of the abnormality degree Sn becomes large, it can be easily judged visually which signal n has a large deviation from the normal state.

As described above, the abnormality degree calculator 103 calculates the abnormality degree for the input data received via the communication I / F 112 at the time of diagnosis processing, and stores these values in the diagnosis result database 110. FIG. 10 is a diagram showing a data structure of the diagnosis result database 110. As shown in FIG. As shown in FIG. 10, the diagnosis result database 110 shows “date and time” column, “abnormality level” column indicating abnormality degree A of the whole plant 5 to be diagnosed, and abnormality contribution degree (Rn) of each signal n. The "abnormality contribution (Rn)" field is stored in a table format. That is, the degree of abnormality A of the entire plant 5 to be diagnosed and the degree of abnormality contribution (Rn) of each signal n are stored in time series.
For example, when the “date and time” is “June 1, 2016 00:00”, “0.01” in the “abnormality degree” column indicating the abnormality degree A of the entire plant 5 to be diagnosed and “abnormality contribution degree” the (Rn) "field stores" a signal a 'of the abnormal contribution R _a "0.0", the abnormal contribution R _B of the "signal B", "0.0". In addition, when the “date and time” is “June 1, 2016 01:30”, “0.05” in the “abnormality degree” column indicating the abnormality degree A of the entire plant 5 to be diagnosed, and “abnormality contribution degree” the (Rn) "field stores" a signal a 'of the abnormal contribution R _a "0.1", the abnormal contribution R _B of the "signal B", "0.5". In the present embodiment, the abnormality contribution degree (Rn) of each signal n is stored, but instead, the abnormality degree (Sn) of each signal n may be stored.

Next, the ART control unit 104 shown in FIG. 1 controls the execution timing of the ART processing unit 105 based on the data of the abnormality degree output from the abnormality degree calculation unit 103.
When the ART processing unit 105 newly creates a category by the processing by ART, it is more efficient to create a new category (abnormal category) when data outside the existing category (normal category) is accumulated. FIG. 11 is an explanatory view of a concept of clustering by ART. In particular, as shown in FIG. 11, it is preferable that a large number of points exist in a collective position in a two-dimensional space. Based on this, the ART control unit 104 activates the ART processing unit 105 to execute the processing by the ART when any of the following conditions is satisfied.
Condition 1: The number of times the degree of abnormality A of the whole plant 5 to be diagnosed exceeds the threshold is a certain number or more Condition 2: the number of times the degree of abnormality Sn of the signal n exceeds the threshold is a certain number or more (counts for each signal) 3: The number of times selected as the nearest normal category is equal to or more than a certain number (counted per category)
The threshold value in each condition is, for example, the calculation result of the abnormality degree A of the whole plant 5 to be diagnosed or the abnormality degree Sn of the signal n in consideration of the fluctuation of the data (characteristics of the data) at the time of the learning process described above. It sets up suitably based on.

  The degree of abnormality A of the whole plant 5 to be diagnosed, the degree of abnormality Sn of the signal n, and the closest normal category are all parameters calculated by the abnormality degree calculator 103. The ART control unit 104 takes in the output of the abnormality degree calculation unit 103 via the internal bus 113, counts the number of times, and determines the execution timing (start timing) of the ART processing unit 105. In the processing after activation of the ART processing unit 105, determination of normality / abnormality is performed in the processing by the above-mentioned ART.

  Next, the display screen of the display device of the input / output device 4 constituting the plant abnormality diagnosis system 1 will be described. FIG. 12 is an example of a display screen that displays the time change of the degree of abnormality and the time change of the category number. As shown in FIG. 12, the display screen 410 of the display device constituting the input / output device 4 is outputted from the first display area 411 that displays the temporal change of the abnormality degree Sn of the signal n, and the ART processing unit 105 It is comprised from the 2nd display area 412 which displays the time change of a category number.

In the example shown in FIG. 12, in the first display area 411, the horizontal axis is date and time, and the vertical axis is an abnormality degree. The abnormality degree S _{A of} the signal _A , the abnormality degree S _{B of} the signal _B , and the abnormality degree of the signal C The temporal change of S _C and the abnormality degree S _D of the signal D are displayed in different colors. The sum of the abnormality degree S _{A of} the signal _A , the abnormality degree S _{B of} the signal _B , the abnormality degree S _{C of} the signal _C , and the abnormality degree S _D of the signal D is the abnormality degree A of the entire plant 5 to be diagnosed. Thus, it is possible for the user to easily grasp visually the breakdown of the temporal change of the abnormality degree A of the whole plant 5 to be diagnosed and the abnormality degree Sn for each signal n. In the example shown in FIG. 12, the abnormality degree A of the whole plant 5 to be diagnosed rapidly increases from around 2:00 on June 1, but it can be seen that the influence of the signal B is large.

  Further, in the example shown in FIG. 12, the second display area 412 has date and time on the horizontal axis and category numbers on the vertical axis, and the time change of the category number output from the ART processing unit 105 described above Display the anomaly category in a distinguishable manner. In the example illustrated in FIG. 12, category numbers 1 to 4 are normal categories, and category numbers (category numbers 5 to 7) exceeding that are category abnormal categories.

  FIG. 13 shows a state in which the category determination processing by the ART processing unit 105 is executed, and the number of plot data of category numbers is increased compared to FIG. As shown in FIG. 13, the time change of the degree of abnormality displayed in the first display area 411 is the same as that in FIG. 12. In the temporal change of the category number displayed in the second display area 412, the number of plot data of the category number 7 is increased by three points as compared with FIG. As described above, the temporal change of the abnormality degree A of the whole plant 5 to be diagnosed displayed in the first display area 411 and the abnormality degree Sn for each signal n are calculated at constant time intervals and periodically displayed on the screen. The time change of the category number displayed in the second display area 412 is displayed at the time when the ART processing unit 105 is executed, while the data accumulated since the previous execution is displayed.

  FIG. 14 is an example of a display screen at the time of registration of an abnormal event. As shown in FIG. 14, the display screen 410 of the display device constituting the input / output device 4 comprises a first display area 411 for displaying the time change of the category number outputted from the ART processing unit 105 and the input / output device 4 It comprises an input area 413 for receiving an input of abnormal event registration by the user via the input device, a registration button 414, and a cancel button 415.

  As shown in FIG. 14, the time change of the category number displayed in the first display area 411 is the same as the display content displayed in the second display area 412 shown in FIG. 13 described above. The input area 413 is an area that enables the user to input an abnormal event, and has an area for selecting and inputting a "category number" and an area for inputting an abnormal event in a text format. In the example illustrated in FIG. 14, of the time changes of the category number output from the ART processing unit 105 displayed in the first display area 411, the input device whose category number is 5 to 7 constitutes the input / output device 4 by the user. It shows the state of selection input via. In addition, "Pump abnormality" is shown in the text input form in the area for inputting an abnormal event. In this screen display state, when the mouse pointer is positioned on the registration button 414 by an operation of a mouse or the like as an input device by the user, and the mouse is clicked, the registration button 414 is activated. Thus, the “category number” column corresponds to “5”, “6”, and “7” in the abnormal event database 111 shown in FIG. 7 described above via the input / output control unit 107 and the internal bus 113. "Pump failure" is stored in the "abnormal event" column. The user may select and input a desired category number from the pull-down menu in the area for selecting and inputting the "category number".

  FIG. 15 is an example of a display screen displaying time changes of category numbers and abnormal events as a diagnosis result. The example illustrated in FIG. 15 illustrates an example of a display screen on “July 15th” one month or more after “June 1st” in which “abnormal event registration” is performed by the above-described user. As indicated by the time change of the category number displayed in the first display area 411, the category number deviates from the normal category at "July 15 12:30", and the ART processing unit 105 sets "5" as the category number. Indicates a state in which the As shown in FIG. 14, since the information on the abnormal event is registered in the abnormal event database 111 for the category number “5” before, the second display area 412 is displayed as “abnormal event”, It is displayed that "Pump abnormality" has occurred as an occurrence event at time "12:30" and that the category number is "5". As described above, if the occurrence abnormality has occurred in the past by using the category number output by the ART processing unit 105, the input / output control unit 107 accesses the abnormality event database 111 through the internal bus 113, Information such as the cause of abnormality stored in the abnormal event database 111 in association with the category number can be provided to the user by displaying it on the screen of the display device constituting the input / output device 4. Further, the user can consider the response operation and the work content while referring to the abnormality information provided by the plant abnormality diagnosis system 1.

  As described above, the plant abnormality diagnosis apparatus 2 constituting the plant abnormality diagnosis system 1 of the present embodiment receives various types of equipment and piping, etc., which are received from the plant control apparatus 3 and which constitute the plant 5 to be diagnosed. Based on a plurality of measurement data (signals) from the sensors, the degree of abnormality A of the entire plant 5 and the degree of abnormality Sn of each signal n are calculated. Thereafter, according to the tendency of the abnormality degree A of the entire plant 5 and / or the abnormality degree Sn of each signal n, the category determination processing is executed by the processing by the ART.

As described above, according to the present embodiment, it is possible to provide a plant abnormality diagnosis device and a plant abnormality diagnosis system capable of early judging an abnormality in the entire plant in abnormality diagnosis of a plant using ART.
Further, according to the present embodiment, the category determination process by ART is not performed at fixed time intervals, and the determination process is executed according to the tendency of the abnormality degree or when data having a high abnormality degree is accumulated. As a result, the calculation load required for the determination process can be reduced, and the total processing time can be shortened, and an abnormality in the entire plant can be determined at an early stage.
Moreover, in addition to the degree of abnormality of the whole plant to be diagnosed, the degree of abnormality for each signal is also displayed on the screen of the display device to visually grasp easily which signal deviates from the normal category it can.
In addition, by associating categories to be determined by ART processing with information about abnormal events and storing them in the abnormal event database, the degree of abnormality of the whole plant to be diagnosed with respect to the same abnormality as an abnormality that occurred in the past By providing the user with information such as an abnormal event in addition to the degree of abnormality after the signal, it is possible to use it for examination of the abnormality response.

The present invention is not limited to the embodiments described above, but includes various modifications.
For example, the embodiments described above are described in detail in order to explain the present invention in an easy-to-understand manner, and are not necessarily limited to those having all the configurations described.

DESCRIPTION OF SYMBOLS 1 ... plant abnormality diagnosis system, 2 ... plant abnormality diagnosis device, 3 ... plant control device, 4 ... input / output device, 5 ... plant, 101 ... data acquisition unit, 102 · · · Data preprocessing unit, 103 · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · Data preprocessing unit, 103 · · · · · · · · · · · · · · · · · · · · ART control unit ... Plant database, 109 ... Category database, 110 ... Diagnosis result database, 111 ... Abnormal event database, 112 ... Communication I / F, 113 ... Internal bus, 210a, 210b ... · Category, 211, 212 · · · Input data, 310 · · · Normalization · noise removal unit (F0 layer), 311 · · · Input pattern holding unit (F1 layer), 312 · · · Category output unit (F2 layer), 313 · · · · · · · Category validity determination unit (Orienting Subsystem), 314 · · · · · · Memory, 410 · · · display screen, 411 · · · the first display area, 412 · · · Second display area, 413 ... input area, 414 ... registration button, 415 ... cancel button

Claims (12)

  Based on the difference in spatial distance between the data belonging to the category determined by the data at the time of normality by the adaptive resonance theory and the plurality of measurement data for a plurality of measurement data from various sensors installed in the plant to be diagnosed A plant abnormality diagnosis apparatus, comprising: an abnormality degree calculation unit that calculates an abnormality degree of the entire plant to be diagnosed. In the plant abnormality diagnosis device according to claim 1,
The abnormality degree calculation unit calculates an abnormality degree for each of the plurality of measurement data based on the difference between the data belonging to the category determined by the data at the normal time and the distance in the axial direction for each of the plurality of measurement data. Plant abnormality diagnostic device characterized by In the plant abnormality diagnosis device according to claim 2,
An ART processing unit that executes category discrimination processing according to adaptive resonance theory;
And an ART control unit that controls the start timing of the ART processing unit,
The ART control unit is characterized in that the ART processing unit is activated when the number of times the abnormality degree of the whole plant to be diagnosed determined by the abnormality degree calculation unit exceeds a predetermined threshold is a predetermined number or more. Plant abnormality diagnosis system. In the plant abnormality diagnosis device according to claim 2,
An ART processing unit that executes category discrimination processing according to adaptive resonance theory;
And an ART control unit that controls the start timing of the ART processing unit,
The ART control unit is characterized in that the ART processing unit is activated when the number of times the abnormality degree of each of the plurality of measurement data obtained by the abnormality degree calculation unit exceeds a predetermined threshold is a predetermined number or more. Plant abnormality diagnosis system. In the plant abnormality diagnosis device according to claim 2,
An ART processing unit that executes category discrimination processing according to adaptive resonance theory;
And an ART control unit that controls the start timing of the ART processing unit,
The plant abnormality diagnosis apparatus, wherein the ART control unit activates the ART processing unit when the number of times the distance to the plurality of measurement data in the space is selected as the closest normal category is equal to or more than a certain number. . A plant abnormality diagnosis apparatus for diagnosing an abnormality of a plant to be diagnosed;
A plant control apparatus for transmitting a plurality of measurement data from process control of a plant to be diagnosed and various sensors installed in the plant to be diagnosed to the plant abnormality diagnosis apparatus;
An input / output device having an input unit and a display unit;
The plant abnormality diagnosis device is a space of data belonging to a category determined by data at normal time by adaptive resonance theory and the plurality of measurement data, with respect to a plurality of measurement data from various sensors transmitted from the plant control device. A plant abnormality diagnosis system, comprising: an abnormality degree calculation unit that calculates an abnormality degree of the entire plant to be diagnosed based on a difference in distance above. In the plant abnormality diagnosis system according to claim 6,
The abnormality degree calculation unit calculates an abnormality degree for each of the plurality of measurement data based on the difference between the data belonging to the category determined by the data at the normal time and the distance in the axial direction for each of the plurality of measurement data. Plant abnormality diagnosis system characterized by In the plant abnormality diagnosis system according to claim 7,
The plant abnormality diagnosis device is
An ART processing unit that executes category discrimination processing according to adaptive resonance theory;
And an ART control unit that controls the start timing of the ART processing unit,
The ART control unit is characterized in that the ART processing unit is activated when the number of times the abnormality degree of the whole plant to be diagnosed determined by the abnormality degree calculation unit exceeds a predetermined threshold is a predetermined number or more. Plant fault diagnosis system. In the plant abnormality diagnosis system according to claim 7,
The plant abnormality diagnosis device is
An ART processing unit that executes category discrimination processing according to adaptive resonance theory;
And an ART control unit that controls the start timing of the ART processing unit,
The ART control unit is characterized in that the ART processing unit is activated when the number of times the abnormality degree of each of the plurality of measurement data obtained by the abnormality degree calculation unit exceeds a predetermined threshold is a predetermined number or more. Plant fault diagnosis system. In the plant abnormality diagnosis system according to claim 7,
The plant abnormality diagnosis device is
An ART processing unit that executes category discrimination processing according to adaptive resonance theory;
And an ART control unit that controls the start timing of the ART processing unit,
The plant abnormality diagnosis system, wherein the ART control unit activates the ART processing unit when the number of times that the distance to the plurality of measurement data in the space is selected as the closest normal category is equal to or more than a certain number. . The plant abnormality diagnosis system according to any one of claims 8 to 10,
The display unit includes a first display area for displaying a temporal change in the degree of abnormality of each of the plurality of measurement data calculated by the degree of abnormality calculation unit, and a category number obtained in the category determination process by the ART processing unit. And a second display region for displaying a time change. The plant abnormality diagnosis system according to any one of claims 8 to 10,
The display unit includes a first display area for displaying a time change of a category number obtained by category discrimination processing by the ART processing unit, and an input area for receiving an input of abnormal event registration via the input unit. A plant abnormality diagnosis system characterized by

Images